WorthIt? 🛍️ — An AI Shopping Assistant That Explains What You're Really Paying For

Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle Submission 💸

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

Buying a product online shouldn't require hours of comparing specifications, watching reviews, and wondering whether a more expensive option is actually worth it.

I built WorthIt?, an open-source AI shopping decision assistant that helps users compare products based on their budget, requirements, and personal priorities.

The idea is simple: instead of just telling users which product scores highest, help them understand why an option fits their needs, what trade-offs they're making, and when a cheaper alternative might be sufficient.

WorthIt? combines a transparent scoring system with an AI assistant that explains product comparisons in everyday language.

The current prototype supports three product categories: wireless earbuds, laptops, and smartphones.

I need wireless earbuds for online classes under ₹1,500. Microphone quality and comfort matter more to me than bass.

WorthIt? uses the available product data to explain which options appear to fit those requirements and what compromises the user might make.

Potential users include students, budget-conscious buyers, and anyone who wants to make more informed purchasing decisions. In the future, the project could explore affiliate partnerships or optional premium features, but these are possibilities rather than existing revenue streams.

One important limitation: the current datasets contain illustrative sample products, not verified live shopping listings. The prototype does not yet fetch real-time prices, stock information, or seller data.

Try WorthIt? — Live Demo Note: WorthIt? currently uses illustrative sample product data rather than verified live shopping listings. The AI assistant explains the available data and helps users understand product trade-offs; it does not independently verify real-world prices, specifications, or availability.

The application is deployed on Render and includes the product comparison interface and AI shopping assistant.

The project is built with Python and Streamlit, with separate modules for scoring, database operations, and AI-generated explanations.

The repository also includes unit tests for the scoring logic and shortlist database. The current test suite has 12 passing tests.

I wanted to avoid making the entire application dependent on an AI model for every decision. Product comparisons involve numerical calculations, and those calculations should be understandable and reproducible.

So I divided the application into two main parts.

The application normalizes supported product features to a five-point scale and calculates weighted match scores using the priorities selected by the user.

It also handles missing feature values so that unavailable information is not automatically treated as a zero score.

This part is implemented in Python using Pandas.

For the language-model component, I integrated the open-weight openai/gpt-oss-20b model through Groq's hosted inference service.

The AI assistant receives the user's written requirements, selected feature priorities, and a limited set of candidate products from the application's existing results.

It then generates a structured comparison that includes a recommendation, a side-by-side table, meaningful differences, and conditional suggestions.

The model is instructed to ground product-related claims in the supplied data and to acknowledge missing information rather than inventing specifications or prices.

The rest of the application uses Streamlit for the interface and SQLite for persistent shortlists.

The application is working, but the AI integration is not infallible. Its explanations can still be incomplete or incorrect, and the current implementation uses hosted inference rather than running the model locally.

For me, open innovation is about being able to inspect, experiment with, and adapt the building blocks of a project instead of treating AI as a black box.

Using an open-weight model gave me the opportunity to explore how a language model can be combined with structured data and explicit scoring logic for a specific problem.

It also lets the project evolve beyond a single generic chatbot. The model is given a defined role: explain the differences between available options, reason about user priorities, and communicate trade-offs.

At the same time, using an open-weight model does not automatically make the entire application independent of external services. WorthIt? currently uses hosted inference through Groq, so it still depends on that service for AI-assisted comparisons.

A future direction would be to evaluate local inference, compare alternative models, and investigate how different models affect recommendation quality, cost, latency, and privacy.

That flexibility is one of the reasons I wanted to build this project around open-source AI.

The current prototype is a starting point, not a finished shopping platform.

Building WorthIt? has been an opportunity to explore how AI can do more than produce a recommendation: it can help explain the reasoning and trade-offs behind a decision.

WorthIt? — Because the best purchase isn't always the cheapest one.